Data-driven dynamic modeling of an electro-hydraulic actuator using ARX system identification

Electro-hydraulic actuators (EHAs) are widely used in industrial and robotic applications due to their high-power density, rapid dynamic response and precise motion capabilities. However, the inherent nonlinearities and parameter uncertainties of hydraulic systems make accurate mathematical modeling challenging. Therefore, this study presents a data-driven system identification approach for modeling an EHA using the AutoRegressive with eXogenous input (ARX) model structure. Experimental input-output data were obtained from an EHA workbench using a multi-sine excitation signal within the frequency range of 0.01 Hz to 1 Hz to sufficiently excite the system dynamics. The collected data were pre-processed and divided into estimation and validation datasets prior to model development using the MATLAB System Identification Toolbox. Several ARX model structures ranging from ARX 221 to ARX 661 were fitted and evaluated to determine the most suitable representation of the system dynamics, where the ARX 331 model achieved the best validation performance with a fit percentage of 84%. The identified model was subsequently represented as a linear discrete-time transfer function and further validated through trajectory tracking analysis between the experimental hardware system and the simulation model. The validation results produced a mean squared error (MSE) of 2.34, a root mean squared error (RMSE) of 1.53 and a fit percentage of 85.83%, demonstrating strong agreement between the experimental and simulated responses. The uniqueness of this study lies in the integration of experimental system identification, transfer function representation and trajectory tracking validation using a practical EHA workbench under multi-frequency excitation conditions. Overall, the findings confirm that the proposed ARX-based data-driven modeling approach is effective in capturing the dominant dynamics of the EHA system and provides a reliable foundation for future modeling and advanced control applications.